What a data center is
A data center is a building, or part of one, designed to house computing equipment: servers that run software, storage systems that hold data, and networking gear that connects everything to the internet. Almost every online service relies on one somewhere, from streaming and email to banking apps and company databases.
They range widely in size. A small server room in an office is a modest version. Large dedicated facilities may hold enormous numbers of servers and are operated by specialist companies or by big technology firms for their own use. When people say a service runs “in the cloud,” they mean it runs on shared equipment in facilities like these. Our cloud computing basics article explains that model from the user’s side.
What is inside
Servers, storage and networking
Servers are specialized computers, typically mounted in tall metal racks. Storage systems hold files and databases. Network switches and routers move data between machines and out to the wider internet. Some servers include accelerators, such as graphics processors, that handle work like machine learning more efficiently than general-purpose chips.
Power systems
Computing equipment needs a steady, clean supply of electricity. Facilities usually have more than one connection to the grid where possible, along with uninterruptible power supplies, which are battery systems that bridge brief interruptions, and backup generators for longer outages. Power conversion and distribution equipment reduces and conditions the incoming electricity, and each conversion step loses a little energy as heat.
Cooling
Nearly all the electricity used by computing equipment ends up as heat. If that heat is not removed, components overheat and fail or slow down. Cooling systems are therefore a core part of the design, not an add-on.
Security and monitoring
Facilities typically control physical access, monitor temperature and power continuously, and have fire detection and suppression systems. These systems also use a small amount of energy.
Why data centers use energy
There are two main reasons, and a third that matters for planning.
The computing itself. Chips use electricity whenever they run. Servers that are switched on but idle usually still draw a meaningful share of their peak power, so a lightly used machine can be an inefficient one.
Cooling and overhead. Moving heat out of the building takes fans, pumps, chillers or other equipment. Power conversion losses and lighting add to this.
Reliability requirements. Many services are expected to be available all the time. To achieve that, operators often build in redundancy, meaning duplicate power paths, spare cooling capacity and extra equipment. Redundancy improves resilience, but it can raise the amount of equipment that must be built and powered.
How cooling works
Approaches vary with climate, cost and the density of the equipment.
- Air cooling. Cold air is delivered to the front of server racks and warm air is removed from the back. Designers often separate hot and cold air streams so they do not mix, which improves efficiency.
- Chilled water systems. Water is cooled by chillers and circulated to units that cool the air.
- Using outside air. In cooler climates, facilities can sometimes bring in outside air or use evaporative techniques instead of running mechanical chillers all the time. Evaporative methods can use significant amounts of water, so water supply is part of the discussion in some regions.
- Liquid cooling. Liquid carries heat much more effectively than air. Some systems circulate liquid directly to the chips or immerse equipment in specialized fluids. This is increasingly considered for dense, high-power hardware.
Each choice involves trade-offs among electricity, water, cost and complexity. There is no single approach that is best everywhere.
Measuring efficiency
A commonly cited measure is power usage effectiveness, or PUE. It is the total energy used by the facility divided by the energy delivered to the computing equipment. A value close to 1 means little energy goes to overhead like cooling; higher values mean more overhead.
PUE is useful but limited. It says nothing about how efficiently the servers use their power, whether they are busy or idle, or how clean the electricity supply is. Two facilities can have the same PUE and very different overall environmental impact. Reported figures are also not always measured the same way, so comparing them carefully matters.
What affects the environmental footprint
Several factors matter:
- Electricity source. A facility powered largely by low-carbon sources has lower emissions than an identical one powered by fossil fuels. Some operators buy renewable energy or sign long-term supply contracts, though how this matches actual hourly consumption is a subject of debate.
- Utilization. Sharing hardware among many customers can raise utilization. Approaches such as serverless computing aim to use capacity more fully, though results depend on the workload.
- Hardware age and efficiency. Newer equipment often does more work per unit of energy, but making it has its own footprint.
- Location and climate. Cooler climates can reduce cooling needs. Location also affects grid mix, land, water and connectivity.
- Waste heat. In some places, heat from facilities is captured and used for nearby buildings, though this requires suitable infrastructure.
Local and policy questions
Large facilities can strain local electricity grids and water supplies, and communities and planners sometimes debate where and how they should be built. Some governments also care about where data is physically stored, which relates to data sovereignty. Another trend is moving some processing closer to users or devices through edge computing, which spreads computing across more, smaller sites rather than concentrating it.
Keeping the numbers in perspective
Data centers are often described as using a large amount of electricity, and in absolute terms that can be true. Whether their share of global demand is large or small, and how fast it is changing, depends on the source, the year and what is counted. Efficiency improvements have at times offset rising demand, and at other times demand growth has outpaced them. Because future demand, particularly from AI workloads, is uncertain, careful readers should be wary of confident forecasts in either direction.
The main takeaway is that a data center is a physical, energy-intensive building whose footprint is shaped by design choices, electricity sources and how well its equipment is used.